Audio Interaction Model
2026-06-03 • Sound
SoundArtificial IntelligenceComputation and LanguageMultimedia
AI summaryⓘ
The authors introduce a new type of audio model called Audio-Interaction that can listen and respond to sounds and instructions in real time, unlike previous models that work offline or handle only one task at a time. They created a system called SoundFlow to make this work smoothly by designing special training and data methods for streaming audio. They also built a large dataset and evaluation tools to test the model's abilities. Their model can do many audio tasks live, such as transcribing speech as it happens and following spoken commands, while keeping up performance on usual tasks.
Large Audio Language ModelsStreaming AudioAutomatic Speech Recognition (ASR)Audio Instruction FollowingPerceive-Decide-Respond LoopStreaming DataLow-Latency InferenceAudio Interaction ModelBenchmarkingProactive Audio Intervention
Authors
Zhifei Xie, Zihang Liu, Ze An, Xiaobin Hu, Yue Liao, Ziyang Ma, Dongchao Yang, Mingbao Lin, Deheng Ye, Shuicheng Yan, Chunyan Miao
Abstract
Audio is an inherently interactive modality, yet today's Large Audio Language Models (LALMs) are offline, and streaming audio models each handle only a single task such as streaming ASR or voice chatting. It is time to unify them into one online LALM: a model that, through an always-on perceive-decide-respond loop, listens to sound, environment, and instructions in real time and reacts on the fly. We formalize this regime as the Audio Interaction Model, and realize it with Audio-Interaction, a unified streaming model that retains offline task execution while adding online general audio instruction following, from dialogue to full voice chatting, deciding when to respond from the semantics of the stream. To enable this, we propose SoundFlow, a framework that instantiates the perceive-decide-respond loop end to end, from data to training to deployment, through streaming-native data construction, comprehension-aware training, and asynchronous low-latency inference for stable real-time interaction. We further construct StreamAudio-2M, a 2.6M-item streaming corpus spanning 7 fundamental abilities and 28 sub-tasks, and Proactive-Sound-Bench for evaluating proactive audio intervention. Across 8 benchmarks, Audio-Interaction preserves competitive performance on mainstream audio tasks while unlocking capabilities inaccessible to offline LALMs, including real-time ASR, streaming audio instruction following, and proactive help.